So, if you're attempting to decrypt encrypted transmissions picked up through classified signals intelligence (are you listening, NSA?), a connected machine using an architecture like computing fabrics might be a big win. But if you're attempting to do one of many other compute-intensive tasks that don't require a dynamically shared memory space, a moderately loosely coupled architecture consisting of networked, high-performance PCs might well do the trick instead.

The search for little green men
For example, the project SETI At Home is attempting to analyze radio telescope signals from the Arecibo radio telescope in Puerto Rico as part of the search for extra-terrestrial intelligence. Whether or not you believe in little green men is not an issue here. What is interesting is that the SETI At Home project plans to use plain ol' PCs that are connected to the Internet on a continuous basis. Project personnel have developed server-based software that divides the radio telescope data into chunks, distributes these chunks to clients, and collects the results. It also manages the distribution of architecture-specific versions of the analysis code, and takes care of various security concerns.

And it all runs as part of a screen-saver. Seriously. So as to not take away from the workaday performance of individual PCs (thereby not interfering with the normal work of the volunteers granting access to their PCs), the analysis occurs only when the PC client is idle, exactly as a screen-saver would work. To make things more fun for the tens of thousands of cycle-contributors, three main "visualizations" of the experiment are available during "screen-saver" time:

• Science Mode shows the analysis taking place on the local machine. The significance of each result is explained, and the process is illustrated to a level understandable by high school science students.

• Sky Progress Mode shows how the entire experiment is covering the sky, and summarizes at a glance all the potentially interesting results found so far. The background for this visualization will usually be an accurate image of the bright stars in that region of the sky, but participants may choose more abstract representations as well.

• Earth Progress Mode focuses on the people participating in the experiment. A view of the earth will be shown with a highlight for every individual or organization who is currently participating. The total number of computers involved will be displayed in real time, and participants who have been involved the longest, or have analyzed the most data, will be featured.

The fluidity of such an architecture is appealing, in the sense that you have the ability to dynamically reconfigure your processing power based on the problems at hand.